Save and load?

[lachesis]

Is there a way to save the scan result and reimport it for interactive view? This would be similar to ncdu's -o and -f options.

[Byron]

That would be interesting for some use-cases for sure and quite straightforward to implement I presume if one uses standard Rust serialization.

[gosuwachu]

@lachesis Can you describe your use case? It would help prfioritize this issue.

[CouldBeThis]

Not that anyone asked me but I do occasionally use pdu's --json-output and --json-input to save and load results. You can produce different visualizations for the same data.

My main simple use case is that I have a bunch of old spinning HDDs connected to a low resource computer. I added HDDs slowly over time as I needed space so there was no scheme of what files were where. Too chaosy. So it was required to learn how space was being used then plan a way to move things around.

• Takes a long time even in rust so in this way I can set a task to collect the data then come back and sort through it later.
• Don't have to keep a terminal/ssh/tmux or any other kind of session active to preserve the information
• If you close a window you don't have to start completely fom scratch as I did on a few occasions.
• The json files can be moved to a better computer which is more pleasant to work on.

It's kind of like using locate instead of find. locate periodically creates a database then lets you very quickly search through it, whereas find starts from scratch every time. If your file-set is relatively stable the speed can be an acceptable trade off.

And since it's json I guess you can do all kinds of things with it.

I think it would be cool if dua could do this because it's the one I like the best but tbh it is also acceptable to use the other available tools. As an intermediary, loading the jsons generated by pdu would be cool because dua does a much superior job of visualizing. But only once in a very long while with a very specific use case.

[Byron]

Definitely CC @piotrwach :).

[piotrwach]

I also have few use cases where having ability to load a file system walk from a file would be very useful:

1. I work with remote machines where I don't want to run dua. I could instead run some shell command that would list all the files that I could then load in dua on my machine. E.g.

$ ssh remote_host find /home -ls > listing.txt
$ dua i --from-file listing.txt

2. I sometimes need to analyse what is taking space in my Amazon S3 bucket. For that I would like to run something similar as in 1) e.g.

$ s3cmd ls --recursive "s3://mybucket/*" > listing.txt
$ dua i --from-file listing.txt

Unfortunately the formats of listing.txt in both cases are quite different:

find:

$ find . -ls
19692507        0 drwxr-xr-x    7 pwach            staff                 224  9 Nov 07:39 .
19692508        8 -rw-r--r--    1 pwach            staff                1000  9 Nov 07:39 ./1000bytes
19692511        0 drwxr-xr-x    3 pwach            staff                  96  9 Nov 07:39 ./empty-dir
19692512        0 -rw-r--r--    1 pwach            staff                   0  9 Nov 07:39 ./empty-dir/.gitkeep

vs

s3cmd

$ s3cmd ls --recursive "s3://mybucket/*"
2019-06-14 17:08         123  s3://mybucket/a/1.png
2019-06-15 17:08         456  s3://mybucket/b/2.png

For what it is worth, I have changed dua to achieve 2) already, but since it is not universal / probably a bit out of scope of this project I think we need more universal way of doing this.